On October 28, 2021, at the end of a keynote that ran for more than an hour, Mark Zuckerberg told the audience at Facebook's Connect conference that the company's name no longer fitted what it wanted to become. "Right now, our brand is so tightly linked to one product," he said, MediaPost reported that day. "But over time, I hope we are seen as a metaverse company." Facebook the app kept its name. Facebook the company became Meta, a name drawn from the metaverse it promised to build, and its Reality Labs division would go on to cost more than $70 billion over six years before the company began to cut it back.

Almost five years later, on Tuesday, September 29, 2026, Amazon did something that looks similar from a distance. At its unBoxed conference in San Francisco, it retired the name Amazon DSP, replaced it with DVA, for display, video and audio, and merged the advertising console and the demand-side platform into a single interface relaunched as the Amazon Ads Agent. Kelly MacLean, vice president of science, engineering and product at Amazon Ads, told AdExchanger's James Hercher that the company now believes it has "the simplest experience in the industry."

The shorthand needs one correction before any comparison can be made. Amazon DSP did not become Amazon Ads Agent. The buying product became DVA, and Ads Agent is the conversational layer that, in MediaPost's account of the launch on September 29, "replaces the former Ads Console and Amazon DSP experience." Nor was the agent new. Amazon introduced Ads Agent on November 11, 2025 as an assistant that built campaign structures from uploaded media plans and translated plain-language questions into Amazon Marketing Cloud queries, with every campaign held until an advertiser approved it. What changed in September is that the agent stopped being a feature and became the front door.

Two companies, two renames, and two questions that follow naturally. Is Amazon's move the shape of the next decade of media buying, or a rebrand that will look as dated in five years as a cartoon avatar in a virtual conference room? And what happens to the fees, which have been Amazon's sharpest weapon against The Trade Desk? This special edition draws on reporting in PPC Land, AdExchanger, Digiday, Adweek, MediaPost, Search Engine Roundtable and PPC News Feed going back to 2017. Its argument is that the two renames are opposites more than twins. Meta named a destination that its customers had to travel to, and they never went. Amazon has named a mechanism that already runs, and asks advertisers only to stop touching the controls. That is a far smaller request, which is why it is more likely to succeed. It is also why the fee question matters more than the name.

A name chosen in a bad month

Facebook's rename did not arrive in calm conditions. On October 20, 2021, MediaPost reported that an "under-siege" Facebook would reveal a corporate rebrand at Connect, one that would tie the company to what the piece called a "so-far-vague metaverse concept." The backdrop was Frances Haugen, the former employee who had leaked internal documents earlier that month and accused the company of putting profits first. Arizona's attorney general had separately raised concerns about human smuggling advertisements. A new umbrella name for Facebook, Instagram and WhatsApp would, the article suggested, help address negative perceptions of the Facebook brand.

The trade press was not persuaded that a name could do that. AdExchanger's daily roundup on October 29 noted that the company planned to trade as MVRS from December 1, compared the move with Google's creation of Alphabet in 2015, and observed that Apple's iOS 14 privacy changes were starting to take their toll. Then it delivered its verdict in seven words: "A name change won't fix Meta's rep." Adweek's David Cohen, reporting the announcement on October 28, wrote that skeptics saw the rebrand as a distraction from the company's many problems. A day later, an Adweek opinion column ran under the headline "A Facebook by Any Other Name Is Still Facebook".

There were defenders. Kevin Chesters, strategy partner at Harbour Collective, argued in Adweek on November 3 that rebrands now accepted as normal were mocked at first, and that the reaction said more about resistance to change than about the decision. "Facebook will still be called Facebook," he pointed out, since the rename applied to the parent and not the app. Allen Adamson, the branding veteran and NYU Stern professor, offered the most durable line of the week in an Adweek piece by Robert Klara on November 4: "The most important thing with a name change is to execute it really well." Search, at least, executed quickly. Barry Schwartz noted on Search Engine Roundtable that Google was ranking Facebook's announcement first for the query "meta" about an hour after it went out.

It would be wrong to read the rename as pure crisis management. Three months earlier, on the July 2021 earnings call, Zuckerberg had already told investors that Facebook would come to be seen as a metaverse company, and PPC Land recorded him on July 30, 2021 describing the metaverse as "the ultimate expression of social technology." The strategy was real. The timing was unlucky. The ticker followed later: META replaced FB on the Nasdaq from June 9, 2022, after Roundhill gave up the symbol it had used for a metaverse exchange-traded fund.

Almost nobody noticed that Amazon had renamed itself nine days before Facebook did. On October 19, 2021, MediaPost reported that Amazon Advertising had quietly become Amazon Ads, and that its annual conference, AdCon, would henceforth be called unBoxed. The headline described the move as following Google and Facebook, meaning Google Ads and Facebook Ads, which shows how unremarkable it was. Amazon's rename changed a label and promised nothing. Facebook's changed a label and promised a world. That distinction is the whole of this story.

The bill for a destination

The bill for the world arrived quickly. On February 3, 2022, MediaPost reported that Meta's shares had fallen 26%, wiping out $237 billion in market value, after Facebook lost 1 million daily active users in North America for the first time. The same results showed that Reality Labs, the division building the metaverse, had lost $10 billion in 2021. David Wehner, then the chief financial officer, put the cost of Apple's privacy changes to 2022 advertising revenue "in the order of" $10 billion. The company had renamed itself after a product that was costing it about as much as a competitor's policy change.

Brands tried the new world, cautiously. Wendy's built a Wendyverse in Horizon Worlds, and Digiday reported on August 16, 2022 that the activation was promoted through tweets, Instagram posts and YouTube videos because visiting it required a Meta Quest headset. Alex Chandler, founder and creative director of the Horizon creator studio Vidyuu, put the platform's audience at "something like 300,000 to half a million visitors." In 2021, Chipotle's Roblox maze had drawn more than 30,000 users within a few minutes.

By the first anniversary, Digiday's Marty Swant could summarise the position in two facts. The company's stock price had been cut in half since the rebrand, and Apple's privacy changes could cost Meta $10 billion in 2022, the same amount it had lost the previous year on metaverse-related spending. Meta's answer at the time was a $1,500 headset, the Quest Pro, and enterprise partnerships. Tom Goodwin, co-founder of the consultancy All We Have Is Now, said his own lack of major uses for virtual reality "shows this is all a bridge too far and in the wrong direction."

The advertising business was wobbling at the same moment. On October 27, 2022, AdExchanger reported that third-quarter ad revenue had fallen 4% to $27.2 billion, with ads served up 17% and the average price per ad down 18%, and that Meta expected to be the same size or slightly smaller by the end of 2023. A month later, Nicola Mendelsohn, who ran Meta's Global Business Group, defended the bet in an interview with AdExchanger's Allison Schiff, saying "We believe the metaverse is five to 10 years out." Reality Labs had lost nearly $3.7 billion in that quarter alone, and Meta had just cut about 11,000 jobs, roughly 13% of its staff.

What followed was not a metaverse year but a "year of efficiency," the theme Zuckerberg set for 2023 as 2022 revenue came in at $116.6 billion, down 1%. By May 2023, MediaPost counted about 21,000 people cut since the previous November, roughly a quarter of the workforce, and noted that Meta had spent $13.7 billion on Reality Labs in 2022. In March of that year, Meta had already wound down its NFT project after less than a year, with Stephane Kasriel, its head of commerce and fintech, describing the move as "winding down digital collectibles (NFTs) for now." Marketers moved too. Adweek reported on May 17, 2023, on a Sitecore survey of 800 marketers in the United States and United Kingdom showing that interest had shifted from the metaverse to generative AI. By November 2024, an Adweek infographic opened with a sentence that would have seemed absurd three years earlier: "After being everywhere in 2021, the metaverse has virtually disappeared."

The formal retreat took longer. "Four years after rebranding itself for this seemingly inevitable future," AdExchanger's roundup on December 5, 2025 reported Meta's plans to cut metaverse spending significantly within Reality Labs, adding that "the ending was obvious from the start." On January 14, 2026, the same roundup was blunter still: "The metaverse is all but dead." Reality Labs had cut spending by 30%, had reportedly lost $73 billion in five years, and had laid off 1,000 people that week. McKinsey, the roundup recalled, had once projected a $5 trillion metaverse market by 2030.

In February 2026, MediaPost's Colin Kirkland reported that Meta was separating Horizon Worlds from the Quest and making it mobile-first, with about 1,500 Reality Labs staff laid off and its own virtual reality game studio closed. Samantha Ryan, vice president of content at Reality Labs, said 86% of the time people spend in its headsets goes to third-party apps. By late September 2026, as Meta prepared $1,299 virtual reality glasses for launch in the spring, MediaPost described the history flatly: "Horizon Worlds failed to onboard players". The cumulative cost depends on who is counting. MediaPost has cited more than $70 billion over six years, AdExchanger $73 billion in five, and a BBC presenter, quoted by PPC Land on October 4, 2026, "$80 billion almost" before the project was dropped, though Meta's filings do not break the figure out. Reality Labs lost $4.03 billion in the first quarter of 2026 alone. Jeff Horwitz, the Reuters technology reporter, offered a line in the same interview that applies to more than one company in this story: "Things are behemoths until they're not."

What the name ended up meaning

Here the obvious verdict, that the rename was a flop, runs into an awkward fact. The company that bears the name is far larger than the one that adopted it. Meta reported 2021 revenue of $117.9 billion, up 37%. For 2025, AdExchanger reported revenue of about $201 billion, with just over $196 billion of it from advertising, and Zuckerberg told analysts on January 28, 2026 that "ads are going to be by far the most important driver of growth" in the business for the next couple of years. In the second quarter of 2026, advertising revenue reached $59.36 billion, up 27%, out of total revenue of $60.80 billion. The average price per ad rose 12% and impressions rose 14%, Adweek reported on July 29.

The engine of that growth is not the metaverse but automation. Meta announced the Advantage brand in March 2022, months after the rename, as a consolidation of its automated campaign tools. The Advantage+ suite's annual revenue run rate went from roughly $20 billion at the end of 2024 to $60 billion in the third quarter of 2025 and above $75 billion in the second quarter of 2026. Advertisers did not have to buy a headset or visit a virtual island to use it. They had to accept defaults, and they did. The parallel with Amazon is the most important one in this piece, and it will return.

The money now flows to a different future. PPC Land reported on August 1, 2025 that Zuckerberg had set out a vision of "personal superintelligence", after a $14.8 billion investment that brought Alexandr Wang to lead Meta Superintelligence Labs. Digiday reported on August 8, 2025 that the new lab was already part of the pitch to marketers, and contrasted the measured rollout with the October 2021 pivot, which was "anything but measured." Meta's 2026 capital expenditure guidance has moved from $115 billion to $135 billion at the start of the year to $130 billion to $145 billion in July, and its head of infrastructure, Santosh Janardhan, said in a video made as a paid partnership with Meta that the company would spend "well over $100 billion this year alone" on AI infrastructure. Free cash flow in the second quarter was $784 million.

The rhetoric has moved with the money. On August 10, 2026, Zuckerberg published a 6,542-word letter titled "The Future is for Everyone." PPC Land counted the words on August 12 and found that "advertising," "Facebook" and "Instagram" each appear zero times, while "balance of power" appears 23 times. Advertising accounted for about 97.6% of the quarter's revenue. This week's announcements show where the AI money meets the ad business. On October 14, PPC Land reported Meta's claim that its GEM ads model lifted conversions by 18.9% and click-through rates by 12%, figures published without a baseline, time period or methodology. A day earlier, Meta said large language models now sit in the ranking pipeline of the Facebook feed, which takes typed requests. And on October 15, PPC Land reported DataDome's finding that Muse, Meta's personal AI agent released in the United States on September 8, accounted for 33% of agentic browser requests in September, routed almost entirely through cloud infrastructure and not identifying itself as an AI agent.

So was the Meta rename a flop? As a statement of strategy, by the company's own retreat, yes. Facebook renamed itself for a destination, poured somewhere between $70 billion and $80 billion into Reality Labs, the division charged with reaching it, and turned back. As a name, it survived precisely because it said so little. "Meta" now describes an advertising company that spends its profits on artificial intelligence, and nobody is asking it to change back. The lesson for Amazon is double-edged. A name can outlive the bet it was coined for. And the bet that matters is the one advertisers fund, which at Meta turned out to be automation, not avatars.

Amazon has renamed its buying tools before

Amazon's history of renaming is longer and less dramatic. On September 5, 2018, AdExchanger reported that the Amazon Advertising Platform had become Amazon DSP, and that Amazon Media Group and Amazon Marketing Services were retired under a single Amazon Advertising name. Digiday's cheat sheet the next day noted that Headline Search Ads had become Sponsored Brands, that Amazon's ad revenue had reached $2.2 billion in the quarter, up 129%, and that Google had dropped the DoubleClick name about two months earlier.

That Google precedent is the closest historical match for what Amazon did in September. On June 27, 2018, Google renamed AdWords to Google Ads after almost 18 years, folded its DoubleClick advertiser products into Google Marketing Platform and its publisher products into Google Ad Manager. Barry Schwartz called it "just a brand change" on Search Engine Roundtable. It was a little more than that. Digiday's Kerry Flynn reported that the same announcement introduced Display & Video 360 as the consolidated buying product and made Smart Campaigns the default for new advertisers. Google renamed its demand-side platform and made automation the default for newcomers in a single move. Eight years later, DV360 is still the name buyers use, which suggests that renames of buying tools tend to stick when the software underneath keeps working.

Amazon's own path to September 2026 ran through three steps in less than a year. At unBoxed in Nashville on November 11, 2025, it merged the Amazon DSP and the Ads Console into a single product called Campaign Manager and launched Ads Agent as an in-platform assistant. "We just constantly hear how complex it is right now," MacLean told AdExchanger at the time. Digiday reported the same day that Sponsored Ads advertisers could run video, audio and display formats in Campaign Manager without added fees, while the DSP kept its minimum spend requirements, and quoted an anonymous buyer who said, "Candidly we're looking to expand our solutions beyond Amazon now." The next day, Digiday's Seb Joseph examined where agencies add value when setup and pacing are automated, and MacLean described the goal as "more of a crystal box instead of a black box."

On February 2, 2026, Amazon opened a beta of its Model Context Protocol server, letting advertisers connect their own language models to its advertising APIs. "Why not give the agent the instruction manual?" asked Paula Despins, vice president of ads measurement at Amazon Ads. PPC Land reported that the server works with Claude, ChatGPT, Amazon Q and Bedrock, and quoted Ari Paparo, chief executive of Marketecture Media, describing ad tech as "one of the most complex categories of software ever invented."

Then came the defaults. On July 22, 2026, Amazon announced ten AI features for its Brand+ campaigns, four of them switched on by default, including automatic deal selection for new streaming TV prospecting lines. PPC Land's analysis on July 26 observed that Amazon had removed a decision buyers used to make. On July 30, Amazon began moving existing DSP accounts into a single global advertiser account across 34 countries. Brand+ itself had come out of beta for all DSP advertisers in January 2025, promising campaigns built "in as few as four clicks". By September, most of what the rename described already existed. What unBoxed did was put a name on it and move the old controls one click further away.

What unBoxed actually changed

The detail matters more than the name. Digiday's Ronan Shields reported on September 29 that DVA+, now the default campaign setup, consolidates Sponsored Display, Sponsored TV and programmatic display, video and audio buying. AdExchanger added that it runs on Performance+, tuned to sales and conversions, or Brand+, tuned to video branding and prospecting. Advertisers in the simplified mode choose an objective, a budget and creative. Buyers who want control over targeting, deals, supply sources, frequency and bidding must switch on a toggle labelled advanced settings. MacLean described the approach as "human-led AI supported."

Full-Funnel Campaigns go further. They combine sponsored ads, display, video and streaming television in one campaign that can extend to open-internet inventory, including audio. Advertisers approve the creative but "cannot select individual channels or specific audiences," Digiday reported, and Amazon offers visibility at the level of formats only. MediaPost added that Full-Funnel Campaigns require a minimum daily budget of $100 and are available to all advertisers in the United States, measured on new-to-brand and long-term sales. "With DVA+, we have brought the power of programmatic to advertisers of all sizes," MacLean told MediaPost's Laurie Sullivan.

PPC Land's report on the four-step DVA+ campaign builder fills in the timetable. The steps are an objective, products or a brand URL, AI-generated creatives and a budget. The builder is in an extended closed beta in the United States, with open beta expected in late October 2026 and general availability in 35 markets targeted for the first quarter of 2027. Billing is by invoice only during the beta, with credit cards planned for the first quarter of 2027. Publisher include and exclude lists exist only in the advanced tier. Amazon's announcement claimed improvements of up to 68% in cost per acquisition and 34% in return on ad spend, none of them specific to DVA+, and promised the same infrastructure, signals and measurement. The subheading on Lauren Johnson's Adweek report summed up the change as "Goodbye Amazon Ad Console, hello Amazon Ads Agent," and the piece reported MacLean's insistence that it is not simply a rebrand.

The agent is also where Amazon now sells new formats. Branded Conversations and Sponsored Prompts, AdExchanger reported, place advertisers inside Amazon's shopping chatbot, steered by Conversational Briefs, with no ability to choose which reviews the AI surfaces. A cost-per-click product asking shoppers to review items with fewer than 1,000 ratings is managed inside Ads Agent, with a US open beta due in late October. On the developer side, MediaPost reported that Amazon's MCP server had attracted more than 700 external connections since February. It is the only adoption figure to surface since the rename.

An assistant, or an agent?

Strip away the name and a fair question remains: is Ads Agent anything more than a chatbot sitting on top of the old DSP? On the evidence so far, mostly not. Every campaign it builds still waits for an advertiser's approval, as it has since November 2025. The AI-assisted planning and forecasting that would let it move from one campaign to setting up the next was, in AdExchanger's account of the launch, still in development. Performance+ and Brand+, the engines that actually optimise the buying, predate the rename, and Amazon's own announcement promised the same infrastructure, signals and measurement as before. The industry's wider instinct points the same way. Jeffrey Hirsch, chief executive of QuantumPath, said in January that the aim was "to automate the workflow, not the buying decisions," as PPC Land reported on January 17, and platforms have kept a human approval at the point of spend while language models handle planning, setup and reporting. What makes the September changes more than cosmetic is not the chatbot. It is the defaults behind it: DVA+ as the starting point, Full-Funnel Campaigns that do not let advertisers choose channels or audiences, and automatic deal selection switched on for new streaming TV lines.

That distinction also answers whether clients would pay for it. Amazon is not asking them to. No outlet reported a price for Ads Agent, and when Amazon merged its console and DSP in November 2025, Sponsored Ads advertisers were told they could use the new video, audio and display formats without added fees. The cost of running the models is absorbed and recovered through media and data, the same logic behind Amazon's low headline DSP fees. The bill lands elsewhere only when advertisers or agencies build agents of their own. Amazon's MCP server lets them connect their own language model, and that model is metered by the token. Draft Digital hit its Claude usage limits several times in one working day while building its agentic stack, PPC Land reported on August 2, and an IAB Europe demonstration in April completed a three-agent task in about 25 minutes for roughly 50 euros in token and credit charges. "Tokens burn on every call," said Lars Postmus, owner of Draft Digital and Draft Media. Ruben Schreurs, chief executive of Ebiquity, said agencies "can't afford to continue to subsidize all those costs", which they have been treating as a client acquisition expense with no agreed model to replace it. Cyd Falkson, senior vice president of strategic accounts at MediaSense, said agencies "don't know how to price and sell their technology."

Demand for agentic buying as a paid product is still small. Michael Barrett, chief executive of Magnite, put the most optimistic forecasts for 2027 agentic ad spend at $600 million to $700 million, which he called "relatively modest" within programmatic as a whole, and a DataBeat report in June found that conventional programmatic buyers held a 13.4% CPM advantage over AI agents in May 2026 data. Set against that, Amazon's choice looks deliberate. Token-metered tools are being rationed because every query shows up on an invoice. Ads Agent never does. Amazon does not need advertisers to buy the agent, only to keep buying the media it routes, and on a business that took in $19.8 billion in a single quarter, whatever the inference bill comes to can be folded into the media.

Future or flop: the evidence so far

The case that this is the future, at least on Amazon, starts with money. Amazon's advertising revenue reached $19.8 billion in the second quarter of 2026, up 26%, and $76.1 billion over twelve months. In the same results, Amazon said Ads Agent's targeting had delivered 8% lower cost per impression and 6% lower cost per acquisition, its own unaudited figures. Full-year 2025 advertising revenue was $68.6 billion, with Prime Video reaching an average of 315 million ad-supported viewers worldwide.

Buyers are spending more. On October 7, Digiday reported that five media buyers saw client Amazon spend up 15% to 40% year on year before Prime Day. Darren D'Altorio, senior vice president of paid media at Wpromote, said "we're looking at 30% growth year on year" for endemic brands, with 40% for non-endemic ones. Tinuiti put DSP-managed spend up 25% to 30% on 2025. "It is becoming increasingly more defensible to the CFO why we should be investing there," D'Altorio said, while Hillary Bram, vice president of performance buying at Exverus by Brainlabs, offered a shorter version: "Full funnel on Amazon is what's working." Supply is following. On October 6, Sky Media made its UK and Irish streaming inventory, across Sky Go, NOW, Sky Glass, Sky Q and Sky Stream, buyable through Amazon Ads Agent, with programmatic guaranteed deals available immediately and no pricing, volumes or measurement disclosed. Guideline data cited by PPC Land on August 7 showed Amazon growing from under 10% to just under 20% of tracked programmatic spend.

The case for caution is about control and proof rather than demand. Adweek's Commerce Advantage report on October 7 found that Amazon's AI-driven future leaves some buyers wary of losing control, and noted that some agencies have built their own agentic tools for managing Amazon. Measurement is the other weak point. Amazon's new long-term sales metrics estimate that roughly one-fifth of a brand's sales value arrives after the attribution window closes, based on more than 2 million campaigns in June, but agencies get only 90 days of historical sales data where a marketing mix model needs about 18 months, Digiday reported on October 1. Radoslav Mazganov, vice president and head of technology at Hyperzon by Moburst, said: "So far, we haven't seen it move the needle." In AdExchanger the same day, Hercher called lifetime value forecasts "still guesses" and noted that the data cannot be exported. No independent adoption or performance data for DVA+ has been published since September 29.

The rename itself carries small costs. Amazon has now changed the name of its main buying interface twice in under a year, from the DSP and console to Campaign Manager in November 2025 and to Ads Agent in September 2026. PPC Land's glossary entry on the preferred programmatic partner pointed out on October 14 that earlier announcements naming Amazon DSP as a preferred partner now name a product label that no longer exists. The clearest example is Microsoft. When Microsoft decided to wind down its Invest DSP by February 28, 2026, PPC News Feed reported on October 12, 2025 that Amazon DSP would be the preferred partner for Microsoft's advertisers, with Microsoft Monetize joining the Amazon Certified Supply Exchange. The DSP those advertisers were sent to now has a different name.

Is there a Twitter here? The rename of Twitter to X in July 2023 is the best-known recent example of a rename that coincided with an advertiser exodus. Digiday reported that advertisers got no warning before the blue bird disappeared, a week after Elon Musk admitted ad revenue was down 50%, and by October 2023 Guideline data showed an average year-on-year decline of 55%, deepening to 61% between May and August. But X's problem was trust in an owner, not the letter on the logo. Meta's problem was a product few people wanted. Amazon's rename asks neither consumers nor advertisers to trust a new owner or adopt a new medium. It asks advertisers to let software make choices they used to make themselves, which is exactly what Meta's advertisers did with Advantage+ and Google's did with Smart Campaigns.

Some observers think the change is overdue. Brian O'Kelley, now chief executive of Apostra, told Digiday on October 13 that "Amazon is acknowledging that a programmatic-only DSP makes little sense outside a web-centric universe," as part of an argument that the old-school demand-side platform is dead. Dan Salmon of NewStreet Research went further on August 11, telling MediaPost that "Amazon DSP may not even be around in five years". The name lasted seven weeks after that remark. Jeff Green, chief executive of The Trade Desk, had made a similar point from the other side in November 2025, arguing to Digiday that zero-fee DSPs are a way to profit from owned inventory. In a sense, Amazon has agreed with him: what it sells is no longer a DSP in the old meaning, but access to its own inventory, data and agents, with the open web attached.

The verdict on the first question, then, is that this is not a Meta-style flop and is unlikely to become one. Meta's rename depended on consumers adopting a new medium. Amazon's depends on advertisers accepting defaults, and the evidence from Advantage+ is that they do when the results hold. The real uncertainty is narrower and more technical. A default that buyers cannot audit is only as good as the price it pays, which leads to the second question.

What about the fees?

Amazon's fee story is the most consistent part of its pitch. In October 2017, Digiday reported that buyers already described Amazon's DSP fees as among the lowest in the market. By November 2024, Digiday wrote that advertisers typically paid 10% of intended spend through the DSP, against 15% for Google and 20% for The Trade Desk, though the managed service option demanded a minimum of $50,000. Amazon's own documentation, summarised by PPC Land on July 18, 2025, set the fully Amazon-managed option at a $50,000 minimum, a 15% service fee and a 10% DSP platform fee.

Then the fees fell further, and fast. In April 2025, a beta tester told AdExchanger that programmatic guaranteed deals carried 0% DSP fees on any Amazon media and 1% on a set of streaming publishers, with Amazon Marketing Cloud and data fees reducing the total savings, while The Trade Desk's take rate had run between 19% and 21% since 2017. Digiday described Amazon courting buyers with loss-leader tactics, including an extra 10% of a new advertiser's video buy, capped at $250,000. By October 2025, buyers told Digiday that Amazon's DSP fees typically ranged between 4% and 8%, sometimes dropping to 1% and occasionally disappearing. Amazon offered at least one agency a free head-to-head test against a rival DSP, covering inventory, technology and measurement costs, Adweek reported on October 24, 2025. PPC Land's December 2025 summary recorded no fee on programmatic guaranteed deals for Amazon-owned media. The 1% figure that buyers repeat has a narrower scope than it is often given. Amazon's own pitch deck, seen by Digiday in May 2025, "touts a 1% tech fee for programmatic guaranteed deals on premium open web inventory." Both numbers describe guaranteed deals, one way of buying on the platform, not the rate on everything that runs through it.

The pressure showed on the other side of the market. Digiday reported in December 2025 that The Trade Desk was loosening its grip on pricing, with an effective platform fee that could run from the mid-teens to roughly 20% when fully loaded. By August 2026, PPC Land reported that Amazon DSP had been placed at roughly 1% to 2% of media spend against a 12% to 15% range typically associated with The Trade Desk, which does not disclose its take rate. The comparison is less like-for-like than it looks: Amazon's low end reflects guaranteed deals, and neither range includes the audience data that advertisers buy on top. Green argued that rivals believe "their cheap fee will compensate for their lack of objectivity," and illustrated the point with an 8% fee against a 4% fee on $1 of media, producing $1.08 against $1.04. The market was not convinced. The Trade Desk guided third-quarter revenue to at least $650 million, about 12% below the prior year, and its shares fell 24% after hours on August 6. "It is not our goal to be the cheapest platform," Green said. Salmon, quoted by AdExchanger in September, described Amazon as "the apex predator" for The Trade Desk's budgets.

Data is the line item that the headline comparisons leave out, and on Amazon it is not free. When Amazon explained how its regulatory advertising fees are calculated, PPC Land reported in August 2024, it described DSP campaign fees as comprising media cost, audience fees and platform fees. The advertiser who reported 0% guaranteed-deal fees to AdExchanger in April 2025 added that analytics and data fees cut back on the total savings. Inside Amazon Marketing Cloud, third-party signals such as Experian's sit behind a subscription tier called Paid Features, PPC Land's explainer notes. The Trade Desk has its own stack of extras, which AdExchanger listed in March 2026 as platform, tech, bid-shading, OpenPath, data and measurement fees. A buyer comparing the two needs the whole invoice, not the platform percentage.

What did unBoxed change about any of this? On the published evidence, nothing. None of the coverage in AdExchanger, Digiday, MediaPost, Adweek or PPC Land reported a fee change for DVA, a price for using Ads Agent, or any change to DSP minimums. The fee-related details that did surface were small: a $100 daily minimum for Full-Funnel Campaigns, invoice-only billing for DVA+ during its beta, and a cost-per-click model for the new review-request ads. Other costs sit around the edges. Amazon made its own paid audience and retail purchase features in Amazon Marketing Cloud free to query until the end of 2026, without saying what they will cost from January 1, 2027. And in May 2026, an agency executive told Digiday that minimums for direct DSP access had been pushed into the "tens of millions", a threshold that leaves smaller buyers dealing through resellers or Amazon's simplified products. So the honest answer to "what about the fees?" begins with the line items: the platform fee is low, lowest of all on guaranteed deals, and has been falling for years; data targeting is charged on top; and Amazon has given no sign of changing either with the rename.

The line items are not the whole price, however, and this is where the comparison with Meta becomes useful. Meta has never charged advertisers a platform fee. Its take is the auction itself, and in the second quarter of 2026 the average price per ad rose 12%. A buyer on Meta pays nothing for the software and everything through the clearing price. Amazon is moving towards the same model from the other direction. Vlad Chubakov, associate director of programmatic at Delve Deeper, put the logic in one sentence to PPC Land in August: "Amazon can slash margins because they own the pipes." Salmon's reasoning in MediaPost was that DSP revenue carries low gross margins for Amazon, which has an incentive to move it into bundled Amazon Web Services work where margins are higher. A low platform fee, and none at all on Amazon's own guaranteed inventory, is affordable when the profit sits in the media Amazon owns, the audience data it charges for and the cloud contracts it sells.

That is why the Federal Trade Commission's case matters to anyone weighing Ads Agent, even though it does not concern the DSP. On August 31, 2026, the FTC and 22 states filed a 181-page complaint in the Western District of Washington, case 2:26-cv-03097, alleging about $20 billion in hidden surcharges on roughly 1.2 million US advertisers, more than 500,000 of them small and mid-sized businesses. The alleged mechanism is a "soft reserve" price added on top of a second-price auction and capped only by the winning bid. The FTC alleges that the share of clicks billed at the advertiser's own bid rose from about 4% in late 2020 to 79.1% in 2024, after years in which advertisers were told they would pay a penny more than the next-highest bidder. Amazon disputes the case, says advertisers saved more than $8 billion between 2021 and 2025, says about 92% of selected Sponsored Products ads in 2024 were not the highest bid, and insists that "in no scenario does an advertiser pay more than their bid." No liability has been found.

The case covers Sponsored Products, Sponsored Brands and Display Ads. AdExchanger's Anthony Vargas noted on September 3 that the alleged behaviour applies to ads bought through the console, not to Amazon DSP. Sarah Caputo, founder of the buy-side consultancy Fraction Method, examined the DSP side in an AdExchanger column on September 9 and found that Amazon's documentation describes its invitation-only Private Auction product, including Prime Video inventory, as clearing at a "variable CPM based on optimization," with no buyer-visible floor or clearing formula. "Platforms built the systems, and only platforms can fully explain how they set the price," she wrote. In Digiday's account of the case on September 8, an agency executive said buyers were "probably the frogs in the boiling water", and an agency executive said auction type is not passed in log-level data roughly 70% of the time.

Now place those facts beside the unBoxed products. DVA+ merges Sponsored Display, a display product previously bought through the console, with programmatic display, video and audio bought through what used to be the DSP. Full-Funnel Campaigns merge sponsored ads, the products at the centre of the complaint, with DSP and streaming inventory in a single budget that advertisers cannot split by channel. The boundary between a low, published DSP fee and an auction whose pricing is being litigated once ran between two products with separate workflows and sign-ins. It now runs inside one campaign, managed by one agent. When Digiday asked at unBoxed whether the new tools would show Sponsored Ads buyers how individual winning prices were set, including whether a reserve price was involved, Amazon did not commit to providing that level of auction transparency. Amit Bhattacharyya, vice president of full-funnel products, AI and modeling, said only: "We are figuring out how to give them more transparency on the audience side." Digiday's own advice to marketers was to check whether fees, supply sources and auction mechanics remain clear as the buying systems converge.

Meta's European surcharges are sometimes counted as fees, but they are taxes. From July 1, 2026, Meta began adding what it calls location fees of 2% to 5% to advertising served in Austria, Turkey, France, Italy, Spain and the United Kingdom, passing on those countries' digital services taxes. A $100 campaign aimed at Austria costs $105 plus VAT. MediaPost reported that the charge sits on top of budgets, with VAT calculated on the total, and Madison and Wall estimated that the six markets made up about 13% of Meta's 2025 revenue. It is a tax pass-through, not a price for anything Meta provides, and it matters to buyers mainly because it is applied outside Meta's automated optimisation, so real return on ad spend will be lower than dashboards show. "It makes forecasting really challenging," said Luke Knight, co-founder of Prospect Knight. Meta was the last of the three to do this. Google began passing on such taxes in November 2020, when Search Engine Roundtable reported surcharges of 2% in the UK and 5% in Turkey, and Amazon has applied "Regulatory Advertising Fees" since at least August 2024, calculated for DSP campaigns on a base that, PPC Land reported, includes media cost, audience fees and platform fees. Strip out the taxes and Meta's position on fees is simple: it has none. Its take is the auction.

The pattern across both companies is the same. The visible fee, where there is one, is the part that gets negotiated and reported. The larger part of the price is set inside an auction and an allocation system that only the platform can see. Meta has shown that a closed system with no platform fee can grow into an advertising business that Emarketer, in figures reported by Adweek in April, projects will take $243.46 billion in 2026, ahead of Google's $239.54 billion. Amazon is building the same kind of system, with headline platform fees well below The Trade Desk's, its own retail and streaming data, and an agent that decides where the money goes.

The name is the least of it

Five years on, the Meta rename reads as a lesson in the difference between what a name promises and what a company sells. Zuckerberg hoped the company would be seen as a metaverse company. Instead it became an advertising company that funds artificial intelligence, and its founder's 6,542-word letter in August did not mention advertising once. The name survived because it was elastic enough to hold whatever came next.

Amazon's rename is the reverse. It is precise, it describes software that already runs, and it carries no promise that consumers must keep. That makes it more likely to last, and it is the reason the answer to the first question is closer to "future" than "flop." Within Amazon, buying through an agent with automated defaults is no longer an experiment; it is becoming the default, with general availability targeted for early 2027 and the old DSP controls sitting behind a toggle. Whether it is the future for the rest of the market depends on something the rename does not settle. Amazon's headline platform fees are among the lowest in the business, and its prices are among the least visible. Allen Adamson's test from 2021, that a name change must be executed really well, still applies, but the execution that matters this time is not the logo. It is whether "advanced settings" remains a working set of controls, and whether Amazon eventually shows buyers how the agent's prices are set. Until it does, one of the cheapest headline fees in programmatic advertising comes bundled with one of its least transparent prices.

By the numbers

  • $75 billion: Run rate that Meta's Advantage+ automated campaigns exceeded in the second quarter of 2026. PPC Land
  • 79.1%: Amazon clicks billed at the advertiser's own bid in 2024, up from 4% in late 2020, FTC alleges. PPC Land
  • 25%: Total fees on media spend for Amazon's fully managed DSP option: a 15% service fee plus 10% platform fee. PPC Land
  • Q1 2027: Target for DVA+ general availability in 35 markets, with US open beta expected in late October. PPC Land
  • 6,542: Words in Zuckerberg's August AI letter; "advertising," "Facebook" and "Instagram" each appear zero times. PPC Land